Store "age breaks" attribute of contact matrices

[njtierney]

It would be useful to store some kind of "age breaks" attribute along with some of these objects, rather than just relying on the numbers or rows or columns, since those could be the same length, but represent different age breaks.

Related to #127

[njtierney]

This also presents an opportunity to consolidate the existing classes in the package.

Many of the existing classes are just there to control the print method - this could potentially be modified/simplified so that they are just one type of class with new attributes listed that then feed into the print method.

Then a smaller lighter class could be applied on top of that, which helps for example when identifying that the list of setting matrices contains model predictions, which can then be used when generating an NGM or similar.

It also means that we can just use two autoplot methods, one for a matrix with age information, and other for a list of matrices.

The trade off, I guess, is that when creating the new list of matrices, you define the description information that is given for printing when defining it, rather than specifying a custom print method for each one.

Here's a summary of the current classes. These first two contain autoplot methods, and the first one containing the prediction information is used in generate_ngm

• "conmat_setting_prediction_matrix"

  • A list of matrices containing the model predicted contact rate between ages in each setting.
  • used at the end of predict_setting_contacts()
  • generate_ngm.conmat_setting_prediction_matrix
  • autoplot.conmat_setting_prediction_matrix
  • proposed new class: conmat_setting_predictions conmat_setting_matrix

• "conmat_prediction_matrix"

  • used in predict_setting_contacts()
  • A single matrix
  • print.conmat_prediction_matrix
  • autoplot.conmat_prediction_matrix
  • proposed new class: conmat_age_matrix

All of these classes just contain the class information so that they can be used for their own print method - these three are just lists of matrices

• "transmission_probability_matrix"

  • Used at the end of get_setting_transmission_matrices()
  • A list of matrices, each matrix containing the relative probability of individuals in a given age group infecting an individual in another age group, for that setting.
  • print.transmission_probability_matrix
  • proposed new class: conmat_setting_matrix

• "setting_vaccination_matrix"

  • Used in apply_vaccination()
  • A list of matrices, each matrix containing the adjusted number of newly infected individuals for age groups. These numbers have been adjusted based on proposed vaccination rates in age groups
  • print.setting_vaccination_matrix
  • proposed new class: conmat_setting_matrix

• "ngm_setting_matrix"

  • A list of matrices, each matrix containing the number of newly infected individuals for a specified age group.
  • used in calculate_ngm()
  • print.predict_setting_contacts
  • proposed new class: conmat_setting_matrix

These two are different, the first is a list of data frames, one for each setting

• "setting_data"
  • A list of data.frames containing the number of contacts between ages in each setting.
  • used in get_polymod_setting_data()
  • print.setting_data
  • not sure if a new class makes sense here

This one is a list of models, one for each setting.

• "setting_contact_model"
  • A list of fitted bam models for each setting. Each bam model predicts the contact rate between ages, for that setting.
  • Used when performing fit_setting_contacts(), which is used in estimate_setting_contacts()
  • print.setting_contact_model
  • not sure if a new class makes sense here

Thinking this through, here is the summary of changes

First stage - replace the single matrix component:

• new conmat_age_matrix class, replaces conmat_prediction_matrix
• knows about its age groups
• has autoplot method
• conmat_prediction_matrix --> conmat_age_matrix

Second stage: simplify the other matrix classes

• new conmat_setting_matrix class

  • A list of matrices
  • has autoplot method
  • when creating with new_conmat_setting_matrix, provide information for printing
  • knows about age groups, which are in each matrix, from conmat_age_matrix

• conmat_setting_prediction_matrix replaced with two new classes:

• conmat_setting_predictions and conmat_setting_matrix

• generate_ngm.conmat_setting_prediction_matrix --> generate_ngm.conmat_setting_predictions

• autoplot.conmat_setting_prediction_matrix --> autoplot.conmat_setting_matrix

• print.conmat_setting_prediction_matrix --> print.conmat_setting_matrix

The following classes will be removed and replaced like so:

• print.conmat_setting_prediction_matrix --> print.conmat_setting_matrix
• print.transmission_probability_matrix --> print.conmat_setting_matrix
• print.setting_vaccination_matrix --> print.conmat_setting_matrix
• print.ngm_setting_matrix --> print.conmat_setting_matrix

[njtierney]

After thinking this through I'm not convinced it is a good idea to simplify the classes - it involves some refactoring that feels a bit complicated at the moment and I'm not really clear on what the benefits are - outside of meaning we can only have fewer autoplot methods, but the solution to this could be to write new autoplot functions for the new classes.